# data
import pandas as pd #import library
import numpy as np
# graph
import plotly.express as px
import plotly.graph_objects as go
# sys
import sys
import os

# directory path
AppPath = sys.path[0]
data_folder = AppPath + '/data'
file = 'data.csv'

# printing the ways
print(data_folder)
print(AppPath)
# to raw data folder
os.chdir(data_folder)
# import the data
df = pd.read_csv(file, sep='~', parse_dates=True)
# converting dtypes
downcast = 'integer'
errors = 'coerce'
# to int
df['id']= pd.to_numeric(df['id'], downcast=downcast, errors=errors)
df['max_id'] = pd.to_numeric(df['max_id'], downcast=downcast, errors=errors)
df['min_id'] = pd.to_numeric(df['min_id'], downcast=downcast, errors=errors)
df['is_end'] = pd.to_numeric(df['is_end'], downcast=downcast, errors=errors)
df['is_warning'] = pd.to_numeric(df['is_warning'], downcast=downcast, errors=errors)
df['time_delta_min'] = df['time_delta_min'].astype('int64',errors='ignore')
# dt
df['date'] = pd.to_datetime(df['date'])
# float

df

df.info()

print('Number of starts of warnings: ',df[df['unique_warning'] == True].shape[0])
print('Numberof ends of warnings: ', df[df['last_end'] == True].shape[0])

# take from 1 march
df_region = df[(df['region'] == 'Київська') & (df['date']>= pd.to_datetime('3/1/2022'))]
# global var
region = 'Kyiv'
last_date = df_region['datetime'].max()

df_region

last_date

# selecting 
warday = df_region[['month_num','month','war_day', 'unique_warning']]
# grouping
warday = warday.groupby(by=['month_num','month','war_day']).sum()
warday = warday.reset_index()
# plotting
warday_line =  px.line(warday, x='war_day', y="unique_warning", color="month",
                      title=f'intensity of warnings by days in {region}')
warday_line.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'unique_warning', 'month', 'month_num']]
# grouping 
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday']).sum()
weekday = weekday.reset_index()
# printing
weekday_bar =  px.bar(weekday, x="weekday", y="unique_warning", color="month", barmode="group", 
                     title=f'The absolute number of warnings per weekday in {region}')
weekday_bar.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'unique_warning', 'month', 'month_num', 'war_day']] # addedwarday
# grouping 1
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday', 'war_day']).sum()
weekday = weekday.reset_index()
# grouping 2
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday']).median()
weekday = weekday.reset_index()
weekday
# printing
weekday_bar =  px.bar(weekday, x="weekday", y="unique_warning", color="month", barmode="group",
                     title=f'median number of warnings per weekday in {region}')
weekday_bar.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'unique_warning', 'month', 'month_num', 'war_day']] # addedwarday
# grouping 1
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday', 'war_day']).sum()
weekday = weekday.reset_index()
# printing
weekday_boxplot = px.box(weekday, x='month' ,y="unique_warning", color="weekday", 
                        title=f'distribution of warnings per weekday per months in {region}')
weekday_boxplot.show()

# selecting 
warday = df_region[['month','war_day', 'time_delta_min']]
# grouping
warday = warday.groupby(by=['war_day', 'month']).sum()
warday = warday.reset_index()
# plotting
warday_line =  px.line(warday, x='war_day', y="time_delta_min", color="month")
warday_line.show()

# selecting 
warday = df_region[['month','war_day', 'time_delta_min']].copy()
# grouping median
warday_median = warday.groupby(by=['war_day', 'month']).median()
warday_median['measurement'] = 'Median'
# grouping mean
warday_mean = warday.groupby(by=['war_day', 'month']).mean()
warday_mean['measurement'] = 'Mean'
# union
warday = pd.concat([warday_median, warday_mean])
warday = warday.reset_index()
# plotting
warday_line_mean =  px.line(warday, x='war_day', y="time_delta_min", color="measurement",
                           title=f'median and mean duration of one warning by days of war in {region}')
warday_line_mean.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'time_delta_min', 'month', 'month_num']] # addedwarday
# grouping 1
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday']).median()
weekday = weekday.reset_index()
# printing
weekday_bar =  px.bar(weekday, x="weekday", y="time_delta_min", color="month", barmode="group",
                     title=f'median time of warnings per weekday in {region}')
weekday_bar.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'time_delta_min', 'month', 'month_num', 'war_day']] # addedwarday
# grouping 1
weekday = weekday.groupby(by=['month_num', 'month','weekday_num', 'weekday', 'war_day']).sum()
weekday = weekday.reset_index()
weekday
# printing
weekday_boxplot = px.box(weekday, x='month' ,y="time_delta_min", color="weekday", 
                        title=f'distribution of time in warnings per weekday per months in {region}')
weekday_boxplot.show()

# selecting
weekday = df_region[['weekday_num', 'weekday', 'time_delta_min', 'month', 'month_num', 'war_day']] # added warday
weekday = weekday[pd.isna(weekday['time_delta_min']) == False].sort_values(by=['month_num', 'month', 'weekday_num', 'weekday'])
# grouping 1
weekday
# printing
weekday_boxplot = px.box(weekday, x='month' ,y="time_delta_min", color="weekday", 
                        title=f'distribution of time of 1 warning per weekday per months {region}')
weekday_boxplot.show()

# selecting
month_dist = df_region[['weekday_num', 'weekday', 'time_delta_min', 'month', 'month_num', 'war_day']] # added warday
month_dist = month_dist[pd.isna(month_dist['time_delta_min']) == False].sort_values(by=['month_num', 'month'])
# printing
hist = px.histogram(month_dist, x="time_delta_min", color="month", nbins=100, marginal="box" , facet_row="month",
                   title=f'histogram of time of 1 warning per months in {region}')
hist.show()

# function of counting the Kaplan Meier
def kaplan_meier(df, duration_column='', event_column=''):
    df = df[df[duration_column] > 0]
    durations = df.sort_values(duration_column)[duration_column].unique()
    # Initialise the table
    columns = ['duration', 'n_at_risk', 'n_events', 
               'survival_probability']
    km = pd.DataFrame(columns=columns, dtype=np.number)
    km = km.append(pd.DataFrame([[0, df.shape[0], 0, 1]], 
                                    columns=columns))
    # Calculate survival probability for each duration
    for i, t in enumerate(durations):
        n = np.sum(df[duration_column]>=t)
        d = np.sum((df[duration_column]==t) & (df[event_column]==True))
        s = (1 - d / n) * km.loc[i, 'survival_probability']
        km = km.append(pd.DataFrame([[t, n, d, s]], 
                                    index=[i+1],
                                    columns=columns))
#         km = km[(pd.isna(km['survival_probability']) == False)]
    return km

# seleting necessary values
# per all time
df_km = df_region[['time_delta_min', 'last_end']]
df_km = df_km[(df_km['last_end'] == True) & (df_km['time_delta_min'] > 0)]
df_km['survival_probability'] = 1
df_km['event'] = 1
df_km.index = range(0, df_km.shape[0])

# per month
df_km_march = df_region[df_region['month']=='March']
df_km_april = df_region[df_region['month']=='April']

# counting the Kaplan Meier probability
km_march = kaplan_meier(df_km_march, duration_column='time_delta_min', event_column='last_end')
km_april = kaplan_meier(df_km_april, duration_column='time_delta_min', event_column='last_end')
# creating the plot
fig = go.Figure()
fig.add_trace(go.Scatter(x=km_march['duration'], 
                         y=km_march['survival_probability'],
                    line_shape='vh', name=f'March'))
fig.add_trace(go.Scatter(x=km_april['duration'], 
                         y=km_april['survival_probability'],
                    line_shape='hv', name=f'April'))
fig.add_trace(go.Scatter(x=np.array([0, km_march['duration'].max()]), y=np.array([0.5, 0.5]), name="median",
                    line_shape='linear', line_color='rgb(0,176,246)'))
# fig.update_layout(legend=dict(y=0.5, traceorder='reversed', font_size=16))
fig.update_layout(title=f'Kaplan Meiers probability of warnings in {region}', 
                  xaxis_title='duration of warnings (mins)',
                  yaxis_title='probaility',)
fig.show()

# selecting
df_regr = df_region[['month_num', 'month', 'war_day','unique_warning', 'time_delta_min']]

# grouping 1
wardays_warnings = df_regr[['month_num', 'month', 
                            'war_day','unique_warning']].groupby(by=['month_num', 'month', 'war_day']).sum()
# grouping 2
wardays_durations = df_regr[['month_num', 'month', 
                            'war_day','time_delta_min']].groupby(by=['month_num', 'month', 'war_day']).sum()
# joinings
df_durations_warnings = pd.merge(left=wardays_warnings, right=wardays_durations,
                                         left_index=True, right_index=True)
df_durations_warnings = df_durations_warnings.reset_index()
df_durations_warnings['duration of warnings per day (mins)'] = df_durations_warnings['time_delta_min']
df_durations_warnings['number of warnings per day'] = df_durations_warnings['unique_warning']
# printing the figure
fig = px.scatter(df_durations_warnings, x='number of warnings per day', y='duration of warnings per day (mins)',
                 color='month', title=f'scatter of durations and number of warnings in {region}')
fig.show()